Artificial intelligence is entering a new phase, and with it comes a new identity challenge.
Over the past two years, generative AI (gen AI) has reshaped digital interactions, accelerated fraud, and redefined how organisations approach risk. Now, agentic AI has emerged as the next stage of technological evolution. It enables systems to act autonomously across digital transactions. It changes how transactions are initiated, how fraud is executed, and how trust must be established. Organisations are entering a world where both legitimate human users and bad actors will increasingly be represented by AI agents.
What is agentic AI and what does it mean for fraud?
Agentic AI refers to artificial intelligence systems that can autonomously plan, act and execute tasks. In fraud, this means both attackers and defenders can operate at machine speed, transforming how fraud is perpetrated, detected, prevented and managed.
Key takeaways
- Agentic AI shifts fraud from isolated events to continuous, automated activity
- AI agents are becoming active participants in digital transactions
- Traditional identity models are no longer sufficient
- Trust must become continuous, contextual and agent-aware
- New frameworks such as Know Your Agent (KYA) are fundamental
From generative AI to agentic AI in fraud
Gen AI has already lowered the barrier to entry for fraud. It enables the rapid creation of synthetic identities, convincing phishing content, and AI-generated documentation, making fraudulent activity harder to detect at scale. Agentic AI takes this further.
These systems introduce autonomy, and the ability to execute multi-step processes without human intervention. In practical terms, gen AI creates content. Agentic AI carries out actions.
For financial institutions and digital commerce, this marks a shift from human-led interactions to machine-mediated ecosystems, where automation becomes an active participant in decision-making and transactions.
The rise of autonomous fraud attacks
Fraud is evolving in parallel with these capabilities. Criminal networks are beginning to deploy autonomous agents capable of orchestrating attacks end to end. This includes creating synthetic identities, navigating verification processes, and executing transactions at scale. These attacks are:
- Autonomous, operating continuously with minimal human input
- Adaptive, learning from controls and adjusting in real time
- Coordinated, managing complex, multi-step fraud workflows
This represents a structural shift. Fraud is no longer episodic. It becomes persistent, scalable, and self-optimising.
At the same time, AI-enabled threats such as deepfakes and synthetic documentation continue to scale rapidly, further challenging traditional detection models.
Agentic commerce and AI-driven transactions
Agentic AI is not only reshaping fraud. It is redefining legitimate transactions. Consumers are beginning to delegate decisions, purchases, and financial interactions to AI agents. This shift toward agentic commerce introduces efficiency, but also new uncertainty.
Organisations must now answer fundamental questions:
- Who is initiating the transaction, a human or an agent?
- Is that agent authorised to act?
- Where does liability sit when something goes wrong?
Traditional identity and authentication frameworks were not designed for this environment. As machines transact directly with systems, distinguishing between legitimate and malicious automation becomes significantly more complex.
The future of digital identity: From KYC to KYA
As agentic interactions increase, identity must evolve. Static credentials and point-in-time verification are no longer sufficient. Instead, identity becomes:
- Continuous, assessed throughout the interaction
- Contextual, informed by behavioural, device, and network signals
- Dynamic, adapting in real time as risk evolves
This shift is driving the emergence of Know Your Agent (KYA), an extension of traditional identity frameworks that enables organisations to:
- Register AI agents within identity systems
- Link them to verified human or business owners
- Define permissions and authorised scope
- Monitor behaviour continuously for anomalies
The challenge is no longer just verifying identity. It is maintaining trust across autonomous, AI-mediated interactions.
Building trust in an agentic AI ecosystem
Organisations need to rethink how trust is established in a world where machines increasingly transact on behalf of humans.
A new trust architecture is emerging, built on:
- Human-to-agent binding, which securely links agents to verified identities
- Agent registries, which validate provenance and permissions in real time
- Continuous trust scoring, which assesses behaviour dynamically across the ecosystem
These capabilities move fraud prevention beyond static controls toward continuous governance of trust across onboarding, authentication, and transaction monitoring.
At the same time, agentic AI strengthens defence. It enables faster detection, richer contextual analysis, and more efficient investigative workflows, helping organisations respond at the same speed as emerging threats.
A defining moment for fraud, identity and trust
As our latest report highlights, organisations are already moving from experimentation to accountable, connected and trusted AI ecosystems, where identity and fraud controls must evolve together.
Agentic AI accelerates this convergence. For organisations, the implication is immediate. The models, controls, and assumptions that defined fraud prevention in a human-centric world are no longer sufficient.
The organisations that adapt early, evolving identity frameworks, strengthening trust signals, and embracing AI-driven defence, will be best positioned to operate with confidence in an increasingly autonomous digital economy.
